You sit down to work on an essay. You start writing. Some part of you is generating — chasing tangents, following associations, throwing material onto the page. Another part of you is editing — cutting, narrowing, deciding what to keep.
You're switching between these two postures constantly. Most writers don't notice. They feel productive when generating and frustrated when editing, or productive when editing and stuck when generating. The two modes feel like the same activity going better or worse.
Michael Deen, drawing on JP Guilford's 1956 cognitive psychology work, names the two as separate cognitive postures. "There are two opposite modes that help you sift through an ideascape: divergence and convergence (coined by JP Guilford in 1956)."1
Different mental machinery. Different rules. Different stages of the essay.
Deen's framings, lifted verbatim:
Divergence: "To diverge is to be open and open-minded, to chase down tangents, to imagine a world of potential in even the stupidest detail, to drop all filters and see how any thing connects to everything."1
Convergence: "To converge is to be a ruthless editor."1
The combined rule: "Through divergence you discover a better center of gravity, and through convergence you let go of anything outside of its pull."1
Divergence finds the center. Convergence prunes around it. Both required. Neither alone produces an essay.
Most writing failure is mode-confusion. Writers try to converge while they're supposed to be diverging — they self-edit during the first draft, killing tangents before the tangents can produce anything generative. Or they try to diverge while they're supposed to be converging — they generate new material when they should be cutting, and the draft balloons past usable.
Deen's reframe: these aren't "good habits" versus "bad habits." They're two cognitive postures, each appropriate at different moments. The mistake is running one when the moment calls for the other.
The early stages of essay-making are divergent. You don't know your Thesis yet. You don't know which Material will turn out to matter. You don't know what shape the essay wants.
"It is natural to fall in love with your own ideas, but you will have to delete a lot more than you write. An essay is not a chronological record of your thinking. Drafts are meant to be shed. Writing lets you explore the edges of your terrain."1
Divergence does the exploration. You write things you'll later cut. The cutting comes later. "Spend as much time reading your draft as you do writing it. If you riff something out in 45 minutes, spend another 45 to analyze the Material on the page."1 But the analysis comes after the riff, not during it.
The cost of trying to converge during divergence: you cut the ideas that would have generated the Thesis. The Thesis can't emerge from a pre-pruned material set.
The later stages of essay-making are convergent. You have your Thesis. You know your Material's shape. The work now is to eliminate what doesn't serve the Thesis.
"We don't want to expand endlessly outwards and keep everything we find. Instead, we want to explore with the goal of discovering the best possible center — our thesis. Once we find it, the trick is to cut everything that doesn't belong in its orbit, and then re-explore from that new vantage point."1
The cycle: explore (divergence) → lock in center (recognition) → cut (convergence) → re-explore from new vantage (divergence again). Multiple cycles per essay. Each cycle sharpens the center.
The cost of trying to diverge during convergence: the draft bloats. New material gets added when old material should be cut. The Thesis blurs. The essay becomes a hydra (see Thesis as Central Organizing Idea for the Hydra Problem).
The divergence/convergence move recurs across the entire framework:
Material gathering: Divergence to collect; convergence to curate. You collect more than you need; you keep the 10-20% that does work for the Thesis.
Thesis discovery: Divergence to explore many possible centers; convergence to commit to one. Multiple thesis candidates compete; one wins; the others get cut (and saved for future essays).
Structure: Divergence to consider many possible sequences; convergence to pick one. Post-its on a wall, re-arrangeable. Eventually the arrangement locks.
Voice: Divergence to try multiple Tone configurations; convergence to settle on the one that serves the material. Mode-switching within an essay is a kind of mini-divergence; the overall coherent identity is convergent.
The pattern: divergent expansion → recognition → convergent contraction → new divergent expansion at the new center.
Standard writing advice oscillates between "just write" (pure divergence) and "edit ruthlessly" (pure convergence) as if either alone would produce the work. Deen's framework rejects this. Both are required, and the skill is knowing which mode the moment calls for.
The writer who only writes (no convergence) ends up with bloated drafts that never close. The writer who only edits (no divergence) ends up unable to start — they're trying to be a ruthless editor of material they haven't generated yet.
The literacy is: at any given moment, am I supposed to be opening or closing? Many writers don't ask. They oscillate by accident.
Pre-pruning: Trying to converge during early drafting. Each sentence is judged before it's allowed onto the page. The fix: separate generation from judgment by time. Generate first; judge tomorrow.
Bloated convergence: Trying to diverge during late revision. New tangents get added when the essay should be tightening. The fix: lock the Thesis. Use it as the razor. Cut, don't add.
Premature commitment: Skipping divergence entirely. You decide on Thesis before you've explored. The Thesis ends up shallow. The fix: explicitly divergent phases — even when you think you know the Thesis, make yourself generate alternatives.
Endless divergence: Avoiding convergence. The essay keeps growing. The fix: deadline. External constraint forces the convergent phase.
Mode-blindness: Not noticing which mode you're in. Writing while editing, editing while writing, doing both poorly. The fix: name the mode. Right now I'm diverging. I will write without judgment for an hour. Or: Right now I'm converging. I will cut without adding for an hour.
Schedule modes: Block divergent writing time and convergent editing time separately. Don't blend them. Different modes; different sessions.
Run divergent phase explicitly: Generate without judgment. Allow tangents. Allow bad ideas. Allow oddities. "To diverge is to be open and open-minded."1
Run convergent phase explicitly: Cut without mercy. Hold the Thesis at the top of your screen. Any paragraph that doesn't serve it goes. "To converge is to be a ruthless editor."1
Alternate cycles: Multiple cycles per essay. Each cycle sharpens. The first cycle's convergence is rougher than the third cycle's.
Test mode-appropriateness in real-time: When stuck, ask which mode you're in and whether it fits the moment. Often the fix is mode-switch, not effort.
Where Deen converges with Guilford's cognitive psychology: Guilford's Structure of Intellect model named divergent vs. convergent thinking as distinct cognitive processes. Educational psychology has built on this distinction for sixty years. Deen ports the cognitive science into writing-craft. The convergence: writing-mode literacy is just cognitive-mode literacy applied to the work of essay-making. The split: Guilford's research was about intelligence testing and creativity assessment; Deen's application is about practice. The implication: writers who don't know their cognitive mode are operating without a self-monitoring tool that cognitive science has documented for decades.
Where Deen extends Paul Graham's "discovery" writing tradition: Graham wrote that the idea of writers starting from a fixed center is an "intellectual hangover from the Middle Ages." Deen quotes this directly (Thesis L258-260). Graham's claim aligns with the divergent-then-convergent rhythm — discovery is divergence; commitment to the discovery is convergence. The split: Graham frames this as a critique of bad pedagogy; Deen frames it as cognitive architecture. Both arrive at the same practical conclusion: don't pre-commit; explore; then lock in.
Where Deen sits next to Dan Wang on writing rhythm: Wang's annual-publishing discipline runs year-long divergence (capture everything throughout the year) followed by ten-day convergence (final assembly sprint). The cycle is long but the architecture is the same. Deen's framework runs the cycle at compressed time-scales (within a single drafting session). The convergence: both writers separate generation from selection by time. The split: Wang's cycle is annual; Deen's is per-draft. Different time-scales; same underlying logic.
Divergence/Convergence as a concept reaches into adjacent vault territory at three specific points.
Eastern Spirituality — Expansion and Contraction as Tantric Rhythm: Tantric metaphysics treats reality as alternating expansion (sṛṣṭi, manifestation) and contraction (saṃhāra, reabsorption). The practitioner cultivates fluency across both — when to expand consciousness outward, when to contract it inward. Deen's divergence/convergence runs the same architecture in writing-craft register. The structural identity: cognitive work requires alternating expansion and contraction, not constant operation in one mode. The split: tantric tradition makes this cosmological; Deen makes it methodological. What this reveals: the alternation-rhythm is psychologically deep and cross-culturally identified. Writing-craft has rediscovered, without the metaphysical frame, what tantric tradition has named as fundamental.
Behavioral Mechanics — Funnel Architecture as Influence Process (sales/marketing): Sales-funnel architecture runs divergent prospecting (cast wide net; many leads) followed by convergent qualification (narrow to high-probability buyers) followed by convergent close (single deal). The pattern is divergence-then-convergence at industrial scale. Deen's writing rhythm runs the same architecture at individual-essay scale. The structural identity is exact. The split: sales optimizes for transaction; essay-craft optimizes for transmission. What this reveals: divergence-then-convergence is a general cognitive-architectural primitive that applies wherever a generative-then-selective process is required.
Cross-Domain — Exploration / Exploitation Tradeoff: Reinforcement learning research formalizes the exploration/exploitation tradeoff — when should an agent explore new options vs. exploit known-good options? The optimal strategy alternates between modes based on uncertainty about the environment. Deen's divergence/convergence is the writing-craft version of this. Divergent phases are exploration (generate options when you don't yet know which works); convergent phases are exploitation (commit to the option that emerged). Writing-mode literacy is structurally identical to a well-studied problem in machine learning and decision theory. Optimal strategies have been characterized mathematically. The implication: the writing skill might be partially derivable from formal decision-theoretic principles.
Creative Practice — Note Capture and Refinement Practice (Wang): Wang's three-layer system (year-long capture → months-long refinement → ten-day assembly) is divergence/convergence at long time-scale. The capture phase is pure divergence (no judgment; all material welcome). The refinement phase is mid-cycle (some convergence on what's worth keeping, but still open to discoveries). The assembly is pure convergence (lock structure, cut everything outside it). Wang demonstrates the principle Deen names — running each mode in its own time-block rather than oscillating moment-to-moment. The handshake reveals operational implementation: write divergent in long stretches; converge in shorter sprints; never blend the two within a single working session.
Synthesis across the four connections: each domain shows the same architectural principle. The cycle is universal across cultures (tantric expansion/contraction), industries (sales funnels), decision theory (exploration/exploitation), and writing practice (Wang's capture/assembly). The cultural-specificity is in how the cycle is named and ritualized, not in whether the cycle operates. Writers who think they're being undisciplined when they switch between generation and editing are running a universal cognitive architecture in real-time. The discipline isn't to stop switching. The discipline is to switch deliberately — knowing which mode is appropriate to the moment. Mode-blindness is what produces unproductive oscillation; mode-awareness is what produces sustained output.
The Sharpest Implication: If divergence and convergence are distinct cognitive postures with different appropriate moments, then most writers are operating mode-blind — switching modes without knowing they're switching, doing each mode poorly because they're trying to do both at once. The fix is explicit mode-naming. Name which mode you're in. Schedule modes to specific times. Don't let one mode bleed into the other's territory. The compounding effect: mode-discipline produces both stronger drafts (divergence isn't pre-pruned) and stronger revisions (convergence isn't diluted by new tangents). The skill is purely procedural — it requires no new ideas, only the discipline to separate two cognitive functions that are typically blended.
Generative Questions: